Williams appears to be off by a factor of several hundred on this figure alone:
"In general, the production of ethanol consumes roughly four gallons of water for every one gallon of ethanol produced, although the figure ranges from 3.5 to 6 in Minnesota ethanol plants. This figure varies from plant to plant depending on the processes that each plant uses to produce the fuel. In general, newer plants have been growing more efficient with their water use." - link
I don't know the source, but it has been quoted enough. Subsidized, corn derived ethanol doesn't appear to make much sense, IMHO.
I believe he was refering to the amount of water that it takes to grow the plant also. That would have to be taken into accouint as well.